Weather Shocks and Labor Allocation: Evidence from Northeastern Brazil
Yüklə 306,73 Kb.
|
- Bu səhifədəki naviqasiya:
- Weather Shocks and Labor Allocation: Evidence from Northeastern Brazil Área: Economia Agrícola Abstract
- Keywords
- Palavras-chave
- 2. The Brazilian Northeast
- 3. A model of rural household labor
- 4. Data and Empirical Strategy 4.1. Household data
|
Weather Shocks and Labor Allocation: Evidence from Northeastern Brazil1 Área: Economia Agrícola Danyelle Branco Phd student at Universidade Federal de Viçosa Email: danyellebranco@gmail.com (31) 9 8670-6541 José Féres Pesquisador do Instituto de Pesquisa Econômica Aplicada (IPEA) jose.feres@ipea.gov.br
This paper analyzes whether rural households use labor allocation to mitigate the effect of drought shocks in the Northeaster Brazilian context. We first document that water scarcity leads to lower income derived from farm work as main, and higher income from secondary jobs. We then examine the extent to which extreme droughts affect time labor allocation. Our results indicate that an additional drought shock per year is associated with greater likelihood of have more than one job, lower share of farm activities on the total hours worked, and higher share of secondary job. The effects are higher for poorer municipalities. These findings are consistent with a response to reduced agricultural profitability due to water scarcity and show the importance of non-agricultural activities as an autonomous mitigation mechanism. Keywords: Drought shocks; rural households; labor allocation; Northeastern Brazil. Resumo Este artigo analisa se as famílias rurais usam a alocação de trabalho para mitigar o efeito de um choque de seca no Nordeste do Brasil. Primeiramente, fornecemos evidência de que a escassez de água está associada a uma menor renda derivada do trabalho principal, sendo este agrícola ou não, e positivamente relacionada a maior renda de empregos secundários. Em seguida, achamos que choques negativos de chuva estão fortemente correlacionados com as decisões de alocação de mão de obra. Um choque de seca a mais por ano está associado a maior probabilidade de ter mais de um emprego, menor participação do trabalho agrícola no total de horas trabalhadas e maior participação do trabalho secundário. Os efeitos são maiores para os municípios mais pobres. Esses resultados são consistentes com uma resposta de mitigação à rentabilidade agrícola reduzida devido à escassez de água. Palavras-chave: Choques de seca; famílias rurais; alocação de trabalho; Nordeste do Brasil. Classificação JEL: O13, O15, Q1, Q54 1. Introduction A consolidated body of research suggests that the incidence of extreme weather events, such as droughts and floods, will rise in the coming century as a result of increased global average temperature (Coates et al., 2014; IPCC, 2013). The economic costs of these climate-related extreme events may be substantial and far-reaching. Much of the discussion in literature has focused on the direct impacts of extreme weather events on health, agriculture, and income.2 However, increasing attention is being paid to the mechanisms underlying these relationships. One intriguing question is whether families adopt loss-income mitigation strategies in response to extreme weather events. While previous studies provide strong evidence that droughts and floods can have an immediate effect on rural income, the extent to which families adjust labor supply to mitigate these effects has been very little investigated.3 Documenting the quantitative importance of these labor supply and other behavioral household responses is crucial for guiding the targeting of policies intended to mitigate the adverse consequences of climate change. Extreme weather events can have in particular important effects on time allocation of labor. In context where irrigation and genetically improved seed are unavailable, rainfall shocks are likely to negatively affect agricultural productivity, most notably causing lower yields of subsistence crops and reduce income from cash crops. As a result, engaging in agricultural activities become less attractive and household should rise the supply of non-agricultural work to hedge against declining agricultural profitability and consumption smoothing. Therefore, non-farm income plays a significant role in rural households by reducing income volatility. Understanding the labor supply responses to weather shocks is particularly relevant in developing countries. Since these countries are located in areas that are warmer, they are expected to experience a disproportionate share of extreme weather events in the future due to climate change. Moreover, these countries have limited social safety nets and weak institutions, so households do not have access to the portfolio of adaptation strategies or avoidance behaviors often available in more developed countries. This paper intends to show the importance of non-agricultural jobs as an autonomous mitigation mechanism. To do so, we provide empirical evidence on the relationship between rainfall shocks and household labor allocation in the Northeast Brazilian context. We believe that focus on Northeastern Brazil provides a compelling setting to investigate this question. First, it is the driest Brazilian region and it has long been subject to harsh climatic conditions, with recurrent events of drought and rising temperatures, leading to further enhance evaporation and reduce water availability (Ab’Sáber, 1999; Marengo, 2009). Second, one of the most populated semi-arid area of the world is localized in Brazilian Northeast, where more than 10 million inhabitants are located in rural areas. For a huge fraction of this population collecting water for consumption, hygiene, and agricultural production is a daily task that demands energy and resources. Lack of adequate access to water also increases the susceptibility to climatic shocks associated with fluctuations in rainfall (Ab’Sáber, 1999; Cirilo, 2008; Insa/MCTI, 2014; Rocha and Soares, 2015). Furthermore, half of all Brazilian rural dwellers and family farming establishments are in Northeast. Almost all of the total area sown in the region is rainfed. Only 2 percent of net area is irrigated.4 Therefore, we would expect rainfall to be an important driver of agricultural productivity and household income. We make use of high frequency gridded information on precipitation and temperature to construct a municipality-by-year weather dataset which then is combined with household microdata by using place and survey month. Our identification strategy exploits variation in rainfall records over time within municipalities, and relies on the assumption that weather shocks, conditional on municipality and year fixed-effects, are not correlated with other latent determinants of labor supply. This identifying assumption is plausible insofar as households are unlikely to anticipate precisely a rainfall shock at a given moment in time and place. We begin our analysis by providing evidence that negative rainfall shocks affect household income in our setting. Although income registries are likely to be subject to considerable measurement error in household surveys, we still observe that drought shocks are significantly associated with lower income derived from the main job. This is especially true when we consider income derived from agricultural activities. Moreover, higher incidence of drought shocks are significantly associated with increased income from secondary jobs, out of agriculture. These results give us confidence that rainfall shocks are in fact an income shifter in our setting. We then explore the extent to which drought shocks affect labor time allocation. We find that negative rainfall shocks are associated with greater likelihood of being employed in more than one job, lower share of farm activities, and higher share of non-agriculture secondary job. We also assess whether these effects vary heterogeneously according with municipality's level of development. The results indicate stronger effects among families residing municipalities with lower per capita income. Taken in their entirety, these results are consistent with a mitigation response to reduced agricultural profits due to water scarcity. A potential identification issue pervading our analysis is migration. What if families migrate away from areas affected by extreme droughts? Empirically this would be problematic only if families that migrate in response to a negative rainfall shock are different from those who do not. To explore this issue, we estimate the main regressions considering only families that live for at least five years in the current municipality. The results are broadly similar compared to our benchmark specification. In addition, when we explore whether rainfall shocks are associated with predetermined individual or household characteristics, we find no evidence that this happen. Thus, selective migration is unlikely to be a major problem. This is consistent with recent work in rural Pakistan that finds no effect of rainfall on the mobility of men or women (Mueller et al., 2014). A small number of papers, focused mostly on reallocation of main job, have examined the relationship between weather and labor allocation. As part of a larger analysis, Jessoe et al. (2017) evaluate the effects of annual fluctuations in weather on employment in rural Mexico. They find no effect of rainfall or temperature shocks on agricultural sector, but show that non-agricultural labor decreases with increases in extreme temperature. Rose (2001) looks at rural Indian farm households to test labor supply responses to rainfall shock. She finds that the probability of participating in the labor market is significantly greater when unexpectedly low levels of precipitation are faced. To our knowledge, there has been no study of drought shocks on labor allocation as mitigation strategy in Brazil. In this paper, we use more detailed information of the individual's work. We know the number of works the person is employed, whether the individual is employed in agricultural sector or not for each job, and the hours worked in both main and secondary job. The fact we know the hours worked supply and not just whether the person participates or not in the labor market, allow us to look at farm and non-farm work as complementary rather than as substitutes only. We start in the next section with a little contextual information about Northeastern Brazil. In the third section we present our motivating model of the joint rural household decision regarding farm and non-farm labor supply. Section 4 describes our data and empirical strategy. Section 5 presents our benchmark results, and explores further empirical results. Section 6 concludes. 2. The Brazilian Northeast The Brazilian Northeast comprises nine states and 1,794 municipalities. This region is also the poorest and the driest Brazilian region and it has long been subject to harsh climatic conditions, with irregular annual precipitation, recurrent events of drought and rising temperatures. Furthermore, one of the most populated semi-arid area of the world is localized in Northeastern Brazil, where more than 23.5 million inhabitants are located, representing approximately 12% of the country's population (Ab’Sáber, 1999; Marengo, 2009). For a huge fraction of this population collecting water for consumption, hygiene, and agricultural production is a daily task that demands energy and resources. Lack of adequate access to water also increases the susceptibility to climatic shocks associated with fluctuations in rainfall (Ab’Sáber, 1999; Cirilo, 2008; Insa/MCTI, 2014; Rocha and Soares, 2015). In Figure 1, we have yearly precipitation between 1979 and 2016 for the Northeast and for the rest of Brazil. Brazilian average historical precipitation is slightly above 1700 mm. In Northeast the average is quite below than what is observed for the rest of the country (749 mm). The figure illustrates that, in the 38-year interval portrayed, yearly precipitation in Northeastern Brazil did only reach the historical average for the rest of the country at a point in time, which was the year of 1981. The figure also shows the recurrence of rainfall deficits throughout the past decades. The Northeastern Brazil is also the region with vast majority of the rural dwellers, more than 14 million inhabitants, which represents almost 50% of Brazilian rural population. The economy is largely based on extensive agriculture, 73% of rural dwellers have farm work as their principal employment. In this context, 89% of agricultural establishments are classified as family farms, employing more than 6 million people. The majority are small producers (with areas smaller than 10 ha) and occupy less than 5% of agricultural land. Also, almost all of the total area sown in the region is rainfed, with only 2 percent of irrigated net area. In the context of Brazilian Northeast, where most of the farmers have no access to irrigation and genetically improved seed, rainfall shocks can disrupt agricultural production, most notably causing lower yields of subsistence crops and reduce income from cash crops. The limited access to credit or insurance markets and many internal and external constraints and stresses also could affect the farmers choice of mitigation strategies, and the labor market out off agriculture may be an alternative path to help rural households to hedge against declining agricultural profitability and consumption smoothing. Figure1. Yearly precipitation in Northeastern Brazil and in the rest of the country 
Notes: Author's calculation based on data from ERA-Interim, 1979-2016. 3. A model of rural household labor We developed a simple model of the joint rural household decision regarding farm and non-farm labor supply. The household decides to allocate the time T among three activities: leisure (  ), farm labor ( and non-farm labor ( ), such that  . Let c be consumption. The rural household utility function  is concave and twice differentiable. The total utility function of the rural household is
 (1)
Let  denote the non-farm wage and, since the agricultural household is a price-taker in all markets (Singh, Squire and Strauss, 1986), we assume that is determined exogenously. So, household will be paid  for time spent working in non-farm labor. Let  be the revenue from agriculture, which is given by agricultural production. Agricultural production is determined by the amount of farm labor, weather shocks, and fixed capital and land. It may be represented by the production function  , where  is the quantity of labor allocated to farm activities,  the weather shock, and  is capital and land.5 R is a random variable that affects farm profits, a higher value of R indicates better weather. It could be defined as the deviation between the total rainfall in given moment of time and the historical average rainfall.6 Literature shows that rural Northeastern Brazil turns positive rainfall shocks into unequivocally beneficial events, enabling us to assume that  .7 Total rural household income is given by the sum of farm revenue and non-farm income, and it may be represented by  . Thus, consumption will be
 (2)
 (3)
The time allocated to leisure is expressed by  . Thus, we can substitute this into the maximization problem to get
 (4)
We consider the case where rural households allocate time for both farm and non-farm activities. In this case, the first order conditions of the optimization problem (4) are given by  (5)  (6)
where  is the marginal utility of the consume and  is the marginal utility of leisure.
From conditions (5) and (6), one may verify that  (7)
Condition (7) indicates that, on optimum, the farm wage is equal to the wage paid by non-farm activities. To ensure a globally concave objective function, and thus, a unique optimum, we assume that  (8)
We are interested in the effect that weather shocks have on the optimal level of both farm and non-farm labor. That is, how would we expect an adverse weather shock to affect the rural household labor decision? They would use non-farm labor as a mitigation strategy to weather shocks? These questions lead us to our two testable hypothesis:
Proof. From the first order condition:  (9)
Farm work has a positive relationship with R. In other words, an increase (reduction) in R implies in increasing (decreasing) farm labor. In this model, there is only one way that weather shocks affect the choice of farm work. When a drought is faced, the marginal productivity of agricultural labor will reduce, which implies a diminishing in the return of farm labor. Thereby, agricultural activities become less attractive. Household will allocate less time to farm labor, thus reducing
Proof. From the first order condition, we can derivate the effect of weather shocks on the optimal choice of non-farm working:  (10)
Weather shock has two effects on the optimal level of non-farm labor. First, a drought decreases both farmland productivity and the value of agricultural work, which affect the benefit of time spent in farm labor. Thus, non-farm labor becomes more attractive, and household will increase . Second, droughts decreases the value of marginal productivity of farm labor. Since the marginal return associated to non-farm activities is higher, the household could choose non-farm labor above the optimal level, leading to rising the total income, and mitigating the shocks effect.
4. Data and Empirical Strategy 4.1. Household data Our basic source for labor market outcomes in the rural Northeast is from the Brazilian Household Survey (PNAD). Every year since 1967, the Brazilian Bureau of Statistics (IBGE) has implemented the PNAD throughout Brazil during the month of September.8 It is nationally and regionally representative, and contains detailed information on socio-economic and demographic characteristics. Since its implementation in 1967, PNAD passed through many methodological alterations along the years. Thus, we restrict our analysis to the period between 2001 and 2014, for which questionnaires and consistent sampling methodologies were maintained. Importantly for our study, the PNAD asks whether respondents work with agriculture, are self-employed, wage-employee, employers, or whether they grow for their own consumption. In addition, respondents are asked to provide information about the number of jobs they have, and the amount of hours usually spent in each job per week. This allows us to calculate the participation of each job on the total of hours worked. The rural sample is comprised of 145,425 individuals from 40,519 households in 150 municipalities. Employment data are available for 92,006 individuals, among which 47,295 are the head of household.9 We restrict the sample to those living outside of urban areas because our causal factor of interest, rainfall, should mainly have an effect in rural areas. We also restrict the sample to individuals aged 10 to 70. The householder's characteristics, just as gender, age, years of schooling, and family size, are also collected from the PNAD. Our main outcomes of interest include probability of have more than one job, ratio of farm work on the total worked, share of secondary job on the total of hours worked, and likelihood of at least one family member being employed in nonagricultural work (non-farm likelihood). Table 1 presents summary statistics of these variables. Table 1. Summary statistics: rural Northeastern Brazil, 2001-2014.
Notes: This table shows summary statistics from PNAD database. In rural Northeastern, more than 70% of individuals report agriculture as their principal economic activity. Most of them are self-employed, while 25% help another member of the household and do not receive any salary. On average, only seven percent of individuals are employed in more than one job (not shown in the table), and the share of time spent on these secondary occupations of the total hours worked is 2,82%. Yüklə 306,73 Kb. Dostları ilə paylaş:
|